clisp-devel

typedef query

[Compro P. <com...@gm...>]

typedef unsigned long ULONGLONG;
typedef ULONGLONG uint64;
typedef uint64 uintP;

I think uintP refers to pointer as it is used in definition of
"gcv_object_t". But clisp.h doesn't has anything like:
build-MT-debug/lispbibl.c:1164:  typedef unsigned long long  ULONGLONG;

Was that overlooked? Or is it not required?

Re: typedef query

[Bruno H. <br...@cl...>]

Compro Prasad wrote:
> typedef unsigned long ULONGLONG;
> typedef ULONGLONG uint64;
> typedef uint64 uintP;
> 
> I think uintP refers to pointer

Yes, the comment in lispbibl.d says so. Just search for 'uintP'.

> But clisp.h doesn't has anything like:
> build-MT-debug/lispbibl.c:1164:  typedef unsigned long long  ULONGLONG;

In my (non-MT) builds, clisp.h contains

typedef unsigned long ULONGLONG;
...
typedef ULONGLONG uint64;
...
typedef uint64 uintP;

Bruno

Re: typedef query

[Compro P. <com...@gm...>]

According to what I have seen, 32 bit systems have sizeof(long) as 32
bits and on 64 bit systems it is 64 bits. So, why trust "long". Why
not use "int" for 32 bits and "long long" for 64 bits? If there is a
problem then what is it?

On Thu, May 25, 2017 at 5:52 PM, Bruno Haible <br...@cl...> wrote:
> Compro Prasad wrote:
>> typedef unsigned long ULONGLONG;
>> typedef ULONGLONG uint64;
>> typedef uint64 uintP;
>>
>> I think uintP refers to pointer
>
> Yes, the comment in lispbibl.d says so. Just search for 'uintP'.
>
>> But clisp.h doesn't has anything like:
>> build-MT-debug/lispbibl.c:1164:  typedef unsigned long long  ULONGLONG;
>
> In my (non-MT) builds, clisp.h contains
>
> typedef unsigned long ULONGLONG;
> ...
> typedef ULONGLONG uint64;
> ...
> typedef uint64 uintP;
>
> Bruno
>

Re: typedef query

[Bruno H. <br...@cl...>]

Compro Prasad wrote:
> According to what I have seen, 32 bit systems have sizeof(long) as 32
> bits and on 64 bit systems it is 64 bits.

On 64-bit native Windows, 'long' is 32 bits, not 64 bits. That is,
'long' can be smaller than a 'void *'.

> So, why trust "long".

The code trusts the type whose width it has determined. The point of
the 'uint32', 'uint64' types is to use *not* make assumptions about
the width of a specific type such as 'int' or 'long'. Computer architectures
evolve.

> Why not use "int" for 32 bits and "long long" for 64 bits? If there is a
> problem then what is it?

Until recently (2015), you could not use 'long long' portably. Now,
finally, MSVC has this type too.

Bruno

Re: typedef query

[Daniel J. <dan...@gm...>]

int might also be only 16 bit wide. AFAIK both types also could be larger
than 32 respectively 64 bits.

The real solution is to use the fixed width integer types provided by the
(C99 and newer) standard (int32_t and int64_t). I started work on that in
February, but was set back by serious mysterious crashes that appeared even
without any of my changes.

I'll look whether I can polish the changes to a complete patch next week,
that I hopefully can test (assuming CLISP from the repo doesn't crash by
itself now).

Cheers, Daniel

Compro Prasad <com...@gm...> schrieb am Fr., 26. Mai 2017, 07:49:

> According to what I have seen, 32 bit systems have sizeof(long) as 32
> bits and on 64 bit systems it is 64 bits. So, why trust "long". Why
> not use "int" for 32 bits and "long long" for 64 bits? If there is a
> problem then what is it?
>
> On Thu, May 25, 2017 at 5:52 PM, Bruno Haible <br...@cl...> wrote:
> > Compro Prasad wrote:
> >> typedef unsigned long ULONGLONG;
> >> typedef ULONGLONG uint64;
> >> typedef uint64 uintP;
> >>
> >> I think uintP refers to pointer
> >
> > Yes, the comment in lispbibl.d says so. Just search for 'uintP'.
> >
> >> But clisp.h doesn't has anything like:
> >> build-MT-debug/lispbibl.c:1164:  typedef unsigned long long  ULONGLONG;
> >
> > In my (non-MT) builds, clisp.h contains
> >
> > typedef unsigned long ULONGLONG;
> > ...
> > typedef ULONGLONG uint64;
> > ...
> > typedef uint64 uintP;
> >
> > Bruno
> >
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> clisp-devel mailing list
> cli...@li...
> https://lists.sourceforge.net/lists/listinfo/clisp-devel
>

Re: primitive integer types

[Bruno H. <br...@cl...>]

Hi Daniel,

> The real solution is to use the fixed width integer types provided by the
> (C99 and newer) standard (int32_t and int64_t). ...
> 
> I'll look whether I can polish the changes to a complete patch next week,

Will this patch be using gnulib's <stdint.h> replacement that clisp already
makes use of?
https://www.gnu.org/software/gnulib/manual/html_node/stdint_002eh.html

I think the tricky part is to ensure that the generated clisp.h can be used
by any compiler that runs on the same platform and produces code for the
same ABI. (e.g. on Solaris, "gcc -m64" and "cc -xarch=generic64").

Bruno